Professional Documents
Culture Documents
Volume 3 Article 7
2014
Recommended Citation
Keyworth, Paul (2014) "The Acoustic Correlates of Stress-Shifting Suffixes in Native and Nonnative English: An Overview," Linguistic
Portfolios: Vol. 3 , Article 7.
Available at: https://repository.stcloudstate.edu/stcloud_ling/vol3/iss1/7
This Article is brought to you for free and open access by theRepository at St. Cloud State. It has been accepted for inclusion in Linguistic Portfolios by
an authorized editor of theRepository at St. Cloud State. For more information, please contact rswexelbaum@stcloudstate.edu.
Keyworth: Acoustic Correlates of Stress-Shifting Suffixes
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 58
PAUL KEYWORTH
ABSTRACT
Although laboratory phonology techniques have been widely employed to discover the
interplay between the acoustic correlates of English Lexical Stress (ELS) – fundamental
frequency, duration, and intensity - studies on ELS in polysyllabic words are rare, and
cross-linguistic acoustic studies in this area are even rarer. Consequently, the effects of
language experience on L2 lexical stress acquisition are not clear. This investigation of
adult Arabic (Saudi Arabian) and Mandarin (Mainland Chinese) speakers analyzes their
ELS production in tokens with seven different stress-shifting suffixes; i.e., Level 1
[+cyclic] derivations to phonologists. Stress productions are then systematically
analyzed and compared with those of speakers of Midwest American English using the
acoustic phonetic software, Praat. In total, one hundred subjects participated in the
study, spread evenly across the three language groups, and 2,125 vowels in 800
spectrograms were analyzed (excluding stress placement and pronunciation errors).
Nonnative speakers completed a sociometric survey prior to recording so that statistical
sampling techniques could be used to evaluate acquisition of accurate ELS production.
The speech samples of native speakers were analyzed to provide norm values for cross-
reference and to provide insights into the proposed Salience Hierarchy of the Acoustic
Correlates of Stress (SHACS). The results support the notion that a SHACS does exist in
the L1 sound system, and that native-like command of this system through accurate ELS
production can be acquired by proficient L2 learners via increased L2 input. Other
findings raise questions as to the accuracy of standard American English dictionary
pronunciations as well as the generalizability of claims made about the acoustic
properties of tonic accent shift.
1.0 Introduction
It is widely accepted that certain suffixes in English cause a shift in stress in the
root morpheme to the syllable directly preceding the suffix (Celce-Murcia, Brinton, &
Goodwin, 1996; Kreidler, 2004). These stress-shifting suffixes have been labeled Level 1
[+cyclic] suffixes by generative phonologists (Kisparsky, 1982; Halle and Kenstowicz,
1991). Pronunciation experts, including Celce-Murcia et al. (1996), have claimed that the
resultant shift in stress in turn causes a change in the neutralization or vowel reduction in
the unstressed syllable. Koffi (personal communication, September 11, 2012) has
affirmed that these claims about lexical stress shifts have not yet been supported
quantitatively by the subfield of laboratory phonology.
In addition to this concern about validity, although various studies on the acoustic
properties of English word stress do exist, there is a lack of consensus in the literature as
to the relative importance of the acoustic correlates of stress -- fundamental frequency
(F0) (i.e., pitch), duration, intensity, and spectral reduction. Indeed, various contrasting
versions of what the author hereby coins the Salience Hierarchy of the Acoustic
Published by theRepository at St. Cloud State, 2014 1
Linguistic Portfolios, Vol. 3 [2014], Art. 7
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 59
Correlates of Stress (SHACS) have been proposed: F0 > duration > intensity (e.g., Fry,
1955, 1958; Ladefoged, 2003), duration > F0 > intensity (e.g., Adams & Munro, 1978),
and duration > intensity > F0 (e.g., Beckman & Edwards, 1994). The latter have reasoned
that F0 is only a relevant acoustic correlate of stress with regards to sentential pitch
accent.
Furthermore, most studies have not explored the acoustic properties of the full range of
Level 1 [cyclic] suffixes in the lexicon. In fact, studies on English Lexical Stress (ELS) in
polysyllabic words in general have largely been ignored in favor of disyllabic minimal
stress pairs, as in Fry’s original studies (1955, 1958).
This paper aims to validate the widely-held impressionistic assertions in the literature
about the morphophonemic properties Level 1 [+cyclic] suffixes by providing
quantifiable data. Therefore, the current study is based on quantitative acoustic analyses
of the data using laboratory phonology techniques which have the advantage of
“replicability and robustness” (Post & Nolan, 2012, p. 544) if suitable sampling and
statistical methods are employed. In addition, this project investigates the dichotomous
claims made my acoustical phonetics experts about SHACS. To do this, syllabic F0,
duration, and intensity productions are analyzed in Level 1 [+cyclic] derivations by
native speakers of Midwestern American English (MWAE) dialect. Due to limitations of
time, the researcher does not measure the acoustic correlate of vowel quality (i.e., first
and second formants (F1 and F2), which is in accordance with Lieberman’s study (1960).
From a second language acquisition (SLA) research perspective, the other purpose of this
study is to observe whether there is a correlation between exposure to the L2 and/or L1
background and production accuracy of Level 1 [+cyclic] suffix derivations. As Zhang et
al. (2008) have succinctly noted, most research in the area of English lexical stress
“confound the phonological issue of stress placement with the phonetic problem of
native-like stress production” (p. 4498). Thus, production accuracy here refers to a
twofold distinction: 1) L2 knowledge of where to place the stress in derived words, and
2) native-like production of the acoustic correlates of stress. More specifically, this study
https://repository.stcloudstate.edu/stcloud_ling/vol3/iss1/7 2
Keyworth: Acoustic Correlates of Stress-Shifting Suffixes
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 60
Ancillary findings also raise questions as to the accuracy of IPA pronunciations of certain
Level 1 [+cyclic] derivations provided by Standard American English (SAE) dictionaries
as well as the generalizability of claims made about the acoustic properties of tonic
accent shift. The latter is a theory proposed by Ladefoged and Johnson (2010, p. 119) that
suggests primary-stressed vowels only differ from secondary-stressed vowels with
regards to an increase in F0, making the syllable [+tonic].
Published by theRepository at St. Cloud State, 2014 3
Linguistic Portfolios, Vol. 3 [2014], Art. 7
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 61
- Paired sample t-tests were used to identify significant differences (p < .05)
between the primary [+tonic, +stress] and secondary stressed [-tonic, +stress]
vowels (P vs. S) in each token.
1
Two ratios used: a) [+tonic]/[-tonic] (P vs. All) b) [+stress, + tonic]/[+stress, -tonic] (P vs. S)
https://repository.stcloudstate.edu/stcloud_ling/vol3/iss1/7 4
Keyworth: Acoustic Correlates of Stress-Shifting Suffixes
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 62
- Mean vocalic relative stress ratios of each factor were submitted to ANOVAs, as
well as Tukey HSD post hoc tests, to see if there were significant differences in
stress production among the language groups. Productions of the stem token,
<history>, were omitted from comparisons.
Table 1: Mean Relative Stress Ratios of Primary [+tonic acc.] to Non Primary Vowels [-tonic acc.]
for the Three Acoustic Correlates by Language Group
Paired sample t-tests revealed that, for each token, at least one correlate had a
significantly different value in the primary-stressed vowel than in the mean of all the
other vowels combined. Notwithstanding, the pentasyllabic token <historicity> was
idiosyncratic as it was only significantly different (negatively) with regards to duration.
Since primary vowels were prominent due to the contrast in one or more acoustic
features, we can conclude that stress-shifts in Level 1 [+cyclic] derivations can be
observed quantitatively, at least in words with fewer than five syllables. Table 2
summarizes the significant findings from all the P vs. All paired-sample t-tests for all
three language groups, so that the reader may have a better overall picture of the ways in
which lexical stress manifests itself in the speech of the different L1 groups.
Published by theRepository at St. Cloud State, 2014 5
Linguistic Portfolios, Vol. 3 [2014], Art. 7
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 63
Table 2: Most Salient Acoustic Correlates of Primary Stressed [+tonic] Vowels in Level 1 [+cyclic]
Derivations per Language Group
ü = salient acoustic correlate û = non-salient acoustic correlate
? = borderline saliency based on a non-significant difference between P and All but a large effect size
https://repository.stcloudstate.edu/stcloud_ling/vol3/iss1/7 6
Keyworth: Acoustic Correlates of Stress-Shifting Suffixes
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 64
Figure 3: Comparative Usage of Intensity as an Acoustic Cue to ELS in Level 1 [+cyclic] Derivations
Paired sample t-tests also revealed that Mandarin English L2 speakers deviated more
from the NES norm in that they tended to use F0 as a salient acoustic cue to ELS more
often (Table 2). Thus, the results seem to concur with earlier studies (e.g., Zhang, Nissen,
& Francis, 2008; Keating and Kuo). In general, SHACS for Mandarin English L2
speakers appears to be:
Published by theRepository at St. Cloud State, 2014 7
Linguistic Portfolios, Vol. 3 [2014], Art. 7
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 65
Figure 4: Comparative Usage of Duration as an Acoustic Cue to ELS in Level 1 [+cyclic] Derivations
Paired sample t-tests revealed that durational contrasts between P and All were not as
numerous as they were for NES although Arabic English L2 speakers also produced
negative duration ratios for <historicity> and <historify> (Table 2). In sum, SHACS for
Arabic L1 NNES appears to be:
https://repository.stcloudstate.edu/stcloud_ling/vol3/iss1/7 8
Keyworth: Acoustic Correlates of Stress-Shifting Suffixes
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 66
Mandarin L1 speakers use a slightly smaller F0 range than NES which seems to confirm
the findings of Li and Shuai (2011).
4.42 Intensity
As reported in 4.31, one-way ANOVAs revealed that Mandarin L1 NNES
significantly under-use intensity contrasts between P and All to emphasize primary stress.
This was somewhat unexpected based on previous studies of Mandarin English L2. Still,
paired sample t-tests revealed that intensity was significantly different in P vs. All in all
but one of the tokens (Table 2). Thus, it seems that although intensity is an important
acoustic cue for Mandarin L1 NNES, they still do not employ it in a native-like manner.
4.43 Duration
As reported in 4.32, one-way ANOVAs revealed that Arabic L1 NNES
significantly under-use durational contrasts between P and All to emphasize primary
stress. This may be due to the fact that Arabic English L2 speakers tend not to reduce
vowels as suggested by Zuraiq and Sereno (2005). Not reducing the unstressed vowels in
a token would certainly result in smaller durational contrasts between the primary
[+tonic] and non-primary [-tonic] vowels. The ANOVA results also support
Bouchhioua’s (2008) study which found that duration is not an important correlate of
lexical stress in Tunisian Arabic as it is in English and negative transfer may lead to non-
native accentedness, if not unintelligibility.
Published by theRepository at St. Cloud State, 2014 9
Linguistic Portfolios, Vol. 3 [2014], Art. 7
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 67
In fact, NES were no better at accurately producing this word than Mandarin speakers.
However, it is important to note that for the two nonsense words (i.e., <historious> and
<historial>), the NES performed much better. It is the researcher’s contention that
although these are not real words, NES were able to use the stress-shifting rules that are
stored in the lexicon. Nevertheless, the smooth curve in Figure 6 shows that a significant
correlation was found (r = - .26, p < .05.) for years of English language study and
frequency of errors. Therefore, the longer learners of English have spent studying the
language (i.e., increased L2 input), the fewer pronunciation and stress-placement errors
they make in stress-suffixed words.
From a psychological viewpoint, this strong correlation may relate to how many years of
English L2 education subjects feel they have received. When L2 exposure (i.e., years of
residence in L2 country) was plotted against number of errors, it did not yield significant
correlations. However, the author posits that the study did not have a large enough range
for this variable. As expected, Arabic and Mandarin L1 NNES produce fewer errors the
https://repository.stcloudstate.edu/stcloud_ling/vol3/iss1/7 10
Keyworth: Acoustic Correlates of Stress-Shifting Suffixes
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 68
higher the level of their English L2 proficiency with a significance of p < 0.00 and a
medium effect size of r = -.05 (Figure 7).
Figure 7: Frequency of Pronunciation and Stress-Placement Errors vs. English Proficiency Level
Published by theRepository at St. Cloud State, 2014 11
Linguistic Portfolios, Vol. 3 [2014], Art. 7
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 69
Figure 8: Difference of Mean Mandarin L1 Speaker Ratio of F0 from Mean Native Speaker Ratio of F0
vs. Years of L2 English Study
Figure 9 shows significant correlations were also found when native-like production of
F0 was correlated with English L2 proficiency level, albeit with a very small effect size (r
= -0.0, p < 0.05).
Figure 9: Difference of Mean Mandarin L1 Speaker Ratio of F0 from Mean Native Speaker Ratio of
Duration vs. English Proficiency Level
Encouragingly, the results suggest that Chinese learners of English are able to overcome
their innate difficulties when producing F0 as an acoustic cue to ELS through increased
L2 study.
https://repository.stcloudstate.edu/stcloud_ling/vol3/iss1/7 12
Keyworth: Acoustic Correlates of Stress-Shifting Suffixes
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 70
Figure 10: Difference of Mean Arabic L1 Speaker Ratio of Duration from Mean Native Speaker Ratio of
Duration vs. Years of L2 English Study
Figure 11: Difference of Mean Arabic L1 speaker Ratio of Duration from Mean Native Speaker Ratio of
Duration vs. English Proficiency Level
Comparing these results with those from the ANOVA (Figure 4), it is proposed that
through increased acquisition of the English language, Saudi learners are also able to
overcome the detrimental effects of negative transfer from their L1 sound system.
4.6 Palatal glide epenthesis in the stress-shifting suffixes: <ian>, <ious>, and
<ial>
The researcher identified a surprisingly high incidence of /j/ epenthesis in the
tokens <historian>, <historious>, and <historial> (Figure 12).
Published by theRepository at St. Cloud State, 2014 13
Linguistic Portfolios, Vol. 3 [2014], Art. 7
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 71
30
Number of Productions 25
20
15
10
0
[hɪstɔ́riəәl] [hɪstɔ́rijəәl] [hɪstɔ́riəәn] [hɪstɔ́rijɛn] [hɪstɔ́riəәs] [hɪstɔ́rijəәs]
historial historian historious
Figure 12: Proportion of Standard American English Dictionary Pronunciations
vs. Proportion of Alternative Pronunciations with Epenthetic Palatal Glide by MWAE NES
https://repository.stcloudstate.edu/stcloud_ling/vol3/iss1/7 14
Keyworth: Acoustic Correlates of Stress-Shifting Suffixes
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 72
<history> û ü ü
<historial> û ü ü
<historian> û ü ü
<historic> û ü ü
<historical> ü ü ü
<historicity> û û û
<historify> û ü ü
<historious> û ü ü
5.0 Conclusion
The results presented in this paper yield insight into several interdependent issues
related to lexical stress in polysyllabic English words containing stress-shifting suffixes.
First and foremost, this study provides support to the view that the acoustic correlates of
Published by theRepository at St. Cloud State, 2014 15
Linguistic Portfolios, Vol. 3 [2014], Art. 7
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 73
stress do indeed have a hierarchy of relative salience, hereinafter named SHACS. The
SHACS proposed here is intensity > duration > fundamental frequency (F0). While this
does not exactly match any of the schemes described in the literature, it most closely
resembles the SHACS postulated by Beckman and Edwards (1994): duration > intensity
> F0. Most likely, SHACS is context dependent. For instance, Fry’s (1955, 1958) notion
of SHACS (i.e., F0 > duration > intensity>) may only be relevant to disyllabic
homographs while intensity may only be the most salient acoustic cue in three and four
syllable words. Clearly, more studies on English lexical stress (ELS) in a wide range of
polysyllabic words are needed to validate this hypothesis. What is certain though is that
relative vocalic stress ratios of the three acoustic cues play an important role in
differentiating lexical stress patterns, and there does appear to be a native-norm for
ordering these acoustic signals. Indeed, the various significant correlations described in
this paper support this notion.
From a second language acquisition perspective, there is good evidence to suggest that
native-like command of the acoustic correlates is attainable for English language learners.
Although speakers with different inherent L1 sound systems encounter different problems
when trying to acquire native-like stress production, it favorably appears that they can
overcome these difficulties through increased input of the L2. Not only do experienced
English language learners produce fewer pronunciation errors, they also produce prosodic
contrasts in a more native-like manner. For instance, although Saudi speakers inherently
under-use duration as acoustic cue to ELS - perhaps by not fully reducing vowels as a
result of L1 transfer from the predictable stress system as suggested by other researchers
(Zuraiq & Sereno, 2005; Altmann, 2006; Bouchhioua, 2008) - they are able to use this
acoustic correlate more accurately as their language skills progress. Similarly, Chinese
learners of English are able to overcome the negative transfer of their tonal system by
producing pitch in a more native-like manner as they advance in their studies.
Furthermore, the present results do not support Ladefoged and Johnson’s (2010) theory
of tonic accent shift. Instead, the results actually suggest the opposite; that is, intensity
and duration appear to be responsible for contrasts between primary and secondary
stressed vowels. It will be interesting to observe whether these findings can be replicated,
and whether the variables of token length (disyllabic vs. polysyllabic words) and/or token
delivery method (read utterances vs. natural speech) have any effect.
https://repository.stcloudstate.edu/stcloud_ling/vol3/iss1/7 16
Keyworth: Acoustic Correlates of Stress-Shifting Suffixes
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 74
South Korea, Singapore, the UK, and the US. He earned a B.Sc. degree from the
University of Kent at Canterbury in Molecular and Cellular Biology in 2001 and attained
the University of Cambridge Certificate in English Language Teaching to Adults
(CELTA) in 2008. His area of research interest is the interface between laboratory
phonology and the sociophonetic aspects of second language acquisition. He intends to
pursue a Ph.D. in the field of acoustic phonetics and speech communication.
E-mail: keyworth.paul@gmail.com LinkedIn: www.linkedin.com/in/keyworth/
Thesis Committee: Dr. Ettien Koffi (Chair), Dr. Michael Schwartz, and Dr. Monica
Devers.
Acknowledgements
The study reported here was supported in part by the Student Research Fund - an internal
grant awarded by the Office of Sponsored Programs at Saint Cloud State University. All
copyright belongs to the Acoustical Society of America (2014).
References
Adams, C., & Munro, R. (1978). In search of the acoustic correlates of stress: fundamental
frequency, amplitude, and duration in the connected utterance of some native and
non-native speakers of English. Phonetica, 35, 125-156.
Altmann, H. (2006). The Perception and Production of Second Language Stress: A Cross-
Linguistic Experimental Study. Dissertation Abstracts International, Section A: The
Humanities And Social Sciences, 67(6), 2133-2134. (2006753642)
Beckman, M. E., & Edwards, J. (1994). Articulatory evidence for differentiating stress
categories. In P. Keating (Ed.), Phonological structure and phonetic form. Papers in
Laboratory Phonology III (pp. 7-33). Cambridge: Cambridge University Press.
Boersma, P., & Weenink, D. (2013). Praat: Doing Phonetics by Computer (Version 5.3.32)
[computer program], < http://www.fon.hum.uva.nl/praat/>. Amsterdam: University
of Amsterdam.
Bouchhioua, N. (2008). Duration as a cue to stress and accent in Tunisian Arabic, native
English, and L2 English. Proceedings from Speech Prosody 2008: The Fourth
International Conference on Speech Prosody. (pp. 535-538). Campinas: Brazil.
Carroll, J.B., Davies, P., & Richman, B. (1971). The American heritage word frequency
book. New York: Houghton Mifflin.
Celce-Murcia, M., Brinton, D., & Goodwin, J. M. (1996). Teaching pronunciation: A
reference for teachers of English to speakers of other languages. Cambridge:
Cambridge University Press.
Cowart, W. (1997). Experimental syntax. Thousand Oaks, Sage.
de Jong, K., & Zawaydeh, B. A. (1999). Stress, duration, and intonation in Arabic word-
level prosody. Journal of Phonetics, 27, 3-22.
Published by theRepository at St. Cloud State, 2014 17
Linguistic Portfolios, Vol. 3 [2014], Art. 7
L i n g u i s t i c P o r t f o l i o s – V o l u m e 3 | 75
Flege, J. E., & Bohn, O. S. (1989). An instrumental study of vowel reduction and stress
placement in Spanish-accented English. Studies in Second Language Acquisition,
11, 35-62.
Fry, D. B. (1955). Duration and intensity as physical correlates of linguistic stress. Journal
of the Acoustical Society of America, 27, 765-768.
Fry, D. B. (1958). Experiments in the perception of stress. Language & Speech, 1:2, 126-
152.
Halle, M., & Kenstowicz, M. (1991). The free element condition and cyclic versus
noncyclic stress. Linguistic Inquiry, 22, 457-501.
Keating, P., & Kuo, G. (2012). Comparison of speaking fundamental frequency in English
and Mandarin. Journal of the Acoustical Society of America, 132, 1050-1060.
Kiparsky, P. (1982). From cyclic phonology to lexical phonology. In H. van der Hulst & N.
Smith (eds.), The Structure of Phonological Representations I, pp. 131-175.
Dordrecht: Foris.
Kreidler, C. W. (2004). The pronunciation of English: A course book, 2nd edition. Malden,
MA: Blackwell.
Ladefoged, P. (2001). A course in phonetics, 4th edition. New York: Thomson-Wadsworth.
Ladefoged, P. (2003). Phonetic data analysis: An introduction to fieldwork and
instrumental techniques. Malden, MA: Blackwell.
Ladefoged, P., & Johnson, K. (2010). A course in phonetics, 6th edition. New York:
Thomson-Wadsworth.
Lee, S., & Cho, M. (2011). An acoustic analysis of stressed and unstressed vowels in
English nouns: A cross-language case study. Korean Journal of English Language
and Linguistics 17, 61-88.
Li, B., & Shuai, L. (2011). Suprasegmental features of Chinese-accented English. Journal
of the Acoustical Society of America, 129, 2453.
Lieberman, P. (1960). Some acoustic correlates of word stress in American English.
Journal of the Acoustical Society of America, 32, 451-454.
Maeda, S. (1976): A characterization of American English intonation. (Doctoral
dissertation). MIT: Cambridge, MA. Retrieved from:
http://hdl.handle.net/1721.1/29189. (edsoai.654895205)
Post, B., & Nolan, F. (2012). Data collection for prosodic analysis of continuous speech
and dialectal varaiation. In A. C. Cohn, C. Fougeron, and M.K. Huffman (Eds.),
The Oxford handbook of laboratory phonology (pp. 538-547). NY: Oxford
University Press.
Ryan, K. (2005). Grid-maker.praat [Praat script]. Available from
http://www.linguistics.ucla.edu/faciliti/facilities/acoustic/praat.html
Yoon, T. (2008). Stress analysis script [Praat script]. Available from
http://web.uvic.ca/~tyoon/resource/vq.praat
Zhang, Y., Nissen. S.L., & Francis, A.L. (2008). Acoustic characteristics of English lexical
stress produced by Mandarin speakers. Journal of the Acoustical Society of
America, 123:6, 4498-4513.
Zuraiq, W. (2006). The Production of Lexical Stress by Native Speakers of Arabic and
English and by Arab Learners of English. Dissertation Abstracts International,
Section A: The Humanities and Social Sciences, 66(12), 4375. (2006751445)
https://repository.stcloudstate.edu/stcloud_ling/vol3/iss1/7 18